I built a 60-question Salesforce Admin practice exam last week by being lazy. I took a big bank of questions, told it to hand me 60 at random, and sat the thing. It felt like the real exam. Same length, same clock, same low-grade dread around question 40.
Then I did the one thing almost nobody does with a practice test: I counted the questions by section. The mock I had just praised for feeling real had spent a third of itself on the corners of the platform the current exam barely visits, while the single fastest-growing section on the blueprint showed up once. Once.
Sixty questions. The wrong sixty.
Here is the part nobody puts on the box: "number of questions" is the figure every practice test advertises and the least useful thing about it. A real Salesforce Admin exam is not 60 questions. It is 60 questions distributed on purpose, and the purpose is written down. Rehearse the count without the distribution and you have rehearsed the length of the exam, not the exam.
The blueprint the exam is built on
So what is the distribution? Salesforce publishes it, and they changed it recently, which is exactly why so much practice material is now testing a shape that no longer exists.
The blueprint refreshed on December 15, 2025 (Salesforce renamed the credential to Platform Administrator in the same move). As of that refresh, the sections and their weights look like this:
| Section | Weight |
|---|---|
| Data and Analytics Management | 17% |
| Configuration and Setup | 15% |
| Object Manager and Lightning App Builder | 15% |
| Automation | 15% |
| Sales and Marketing Applications | 10% |
| Service and Support Applications | 10% |
| Productivity and Collaboration | 10% |
| Agentforce AI | 8% |
A few things jump out if you studied for this a couple of years ago. Data and Analytics Management is now the heaviest single section at 17%; it used to sit mid-pack. Automation is still 15%, but it is Flow Builder's 15% now; Process Builder and Workflow Rules reached end of support at the close of 2025, so questions built around them are dead weight. And there is a section that did not exist before at all: Agentforce AI, worth 8%.
The rest of the shape is stable. You get 60 scored questions plus five unscored ones the graders are trialing, 105 minutes, a $200 fee and $100 a retake, and a pass bar of 68% on the current guide. (A lot of older material still says 65%; if a study site quotes you 65%, check the date on everything else it tells you. And Salesforce does not publish pass rates, so ignore anyone who hands you a precise one.)
Sixty marbles, eight buckets
Percentages are abstract. Let me turn them into the test you would sit.
Take 60 questions and pour them into the eight sections by weight, like dropping sixty marbles into eight buckets sized to the percentages. Round the two fractional ones (17% of 60 is 10.2; 8% is 4.8) and it lands clean, summing to exactly 60:
| Section | Weight | Questions in a real 60-question exam |
|---|---|---|
| Data and Analytics Management | 17% | 10 |
| Configuration and Setup | 15% | 9 |
| Object Manager and Lightning App Builder | 15% | 9 |
| Automation | 15% | 9 |
| Sales and Marketing Applications | 10% | 6 |
| Service and Support Applications | 10% | 6 |
| Productivity and Collaboration | 10% | 6 |
| Agentforce AI | 8% | 5 |
| Total | 100% | 60 |
So that is the target. Ten questions on Data and Analytics; roughly five on Agentforce; nine apiece on Config, Object Manager, and Automation. If your last practice run looked nothing like this table, it was measuring something, but it was not measuring your readiness for the exam in front of you.
Why a random pull lies to you
Back to my lazy mock. Why did it misfire so badly?
Because a flat random draw does not know the blueprint exists. It samples wherever the question bank is deepest, and question banks are deepest wherever their source material was thick. Years of Admin prep were written around Configuration and Setup and Object Manager, because that is what the exam used to weight heavily and what is easy to write questions about. So a bank built on that history tilts that way, and a random 60 inherits the tilt. You get a lovely rehearsal for the 2022 exam.
The sections that suffer are the new and conceptual ones: Data and Analytics, and especially Agentforce, where the source material is still thin and the weight on the real exam is climbing. Those are precisely the questions you most need reps on, and precisely the ones a naive draw starves. The mock feels real because it is the right length. It is the wrong shape.
What a weighted allocation does instead
The fix is not clever; it is just work nobody bothers to do by hand. Hand out the questions the way the exam does, by section weight, instead of drawing them flat.
That is the one rule behind how a session gets built on our free Salesforce Admin practice exam: where the exam guide defines weighted sections, your questions arrive in those same proportions. Ask for 60 and the Data and Analytics bucket gets about ten, Agentforce gets about five, every section is guaranteed at least one, and the total lands exactly on the number you asked for. The eight percent that evaporated from my random pull comes back as five real questions you have to answer.
The other half of the point is the results screen. When you finish, the score that matters is not the single percentage; it is the per-section breakdown, because it maps straight onto that blueprint. Going 85% overall while sitting at 3-for-10 on Data and Analytics is the exact profile of someone about to be surprised on exam day, and the only way to see it coming is a mock that put all ten of those questions in front of you in the first place. Every question also links back to the specific Trailhead page it was generated from, so the gap you find at 10pm is a gap you can close by 10:20.
Where this breaks down
Two things a sales page would leave out.
First, weighting only rescues a test big enough to spread. Ask for 15 questions and the 8% and 10% sections round down to one each or disappear, and you are back to a mock that does not resemble the real distribution. If you want the true shape, run the full 60, or more. A 25-question warm-up is fine for drilling one weak section; it is not a dress rehearsal.
Second, matching the distribution does not make these the real exam items, and I would not trust anyone who implies it does. Every question in the bank is generated from official Trailhead material and then checked by a separate AI reviewer against that same source; that is a real process with a real paper trail, but it is not a pipeline to leaked questions and it never will be. A strong score here is evidence you understand the material, not a guarantee you pass. And if your weak spot is a whole section you have never really learned, more reps are the wrong tool: go read, or buy a good study guide for that one section, then come back and rehearse.
Count the marbles
The number on the box (60 questions, 200 questions, whatever a site is bragging about) tells you how long you will be sitting. It tells you nothing about whether you are practicing the right exam. The distribution does that, and the distribution is public; there is no excuse for a practice test to ignore it.
So before your next attempt, count your last mock by section and hold it against that eight-row table. If it matches, good, you are rehearsing the real thing. If it does not, run a full 60-question set weighted to the current blueprint and find out which bucket has been hiding your weak spot. Better to meet those ten Data and Analytics questions now, on a quiet Tuesday night, than for the first time with the clock running.
Count the marbles. Then go book it.